Efficient self-adjustment places of interest finding by eliminating dynamic access point beacons in public places

Abstract

Existing fingerprint-based places of interest (POIs) finding approaches mainly rely on WiFi signal fingerprints of fixed access points. Due to the prevailing of smartphones and 3G networks, more and more mobile hotspots acting as dynamic access points in public places have hindered the effectiveness of the learning and recognition of POIs. How to effectively eliminate those casually appeared mobile hotspots to improve the accuracy of place finding has recently drawn a lot of research attentions. In this paper, we propose the self-adjustment places of interest finding (SAPFI) to deal with the aforementioned problem. By weighted merging of WiFi beacons in the fingerprints of two similar places, the SAPFI is more robust than the existing place finding approaches such as BeaconPrint, PlaceSence, and SensLoc, especially in detecting public places where mobile hotspots are highly likely presented. The experiment results showed that compared to the existing place finding approaches, the SAPFI has the highest precision of 0.94 and the highest recall of 0.8 in POI discovering. In addition, the SAPFI with the highest F1 score of 0.86 also shows its superiority in place learning and recognition.